Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS20050091468 A1
Publication typeApplication
Application numberUS 10/696,716
Publication dateApr 28, 2005
Filing dateOct 28, 2003
Priority dateOct 28, 2003
Also published asUS7412581, US7877572, US20080313383
Publication number10696716, 696716, US 2005/0091468 A1, US 2005/091468 A1, US 20050091468 A1, US 20050091468A1, US 2005091468 A1, US 2005091468A1, US-A1-20050091468, US-A1-2005091468, US2005/0091468A1, US2005/091468A1, US20050091468 A1, US20050091468A1, US2005091468 A1, US2005091468A1
InventorsToshiyasu Morita, Shumpei Kawasaki
Original AssigneeRenesas Technology America, Inc.
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Processor for virtual machines and method therefor
US 20050091468 A1
Abstract
Apparatus and method are described for a data processing device. The data processor includes features suitable for executing a software virtual machine. The data processor provides an instruction set that supports object-level memory protection suitable for high speed operation. Memory control logic is provided to accommodate a configuration having relatively less random access memory (RAM) as compared to re-programmable, nonvolatile memory, and to improve access to the re-programmable, nonvolatile memory.
Images(20)
Previous page
Next page
Claims(64)
1. A method for transferring data in a data processing unit comprising:
receiving a first address;
if the first address contains a predetermined address component, then producing a second address and accessing a memory location at the second address; and
if the first address does not contain the predetermined address component, then accessing a memory location at the first address,
the first address comprising a plurality of address bits, wherein the predetermined address component comprises less than all of the address bits of the first address,
the first address and the second address having a common address component.
2. The method of claim 1 wherein the predetermined address component comprises the high order byte of the first address.
3. The method of claim 1 further comprising executing a virtual machine in a first page of an address space of the data processing unit, the first address being an address in the first page, wherein the second address is an address in another page of the address space of the data processing unit.
4 The method of claim 1 wherein an address space of the data processing unit comprises a plurality of pages, the first address being an address in a first page, the second address being in an address in a second page that is different from the first page.
5. A data processor comprising a memory management unit for handling addresses, the memory management unit comprising:
address input lines for presenting an input address;
address translation logic coupled to at least some of the address input lines; and
address output lines coupled to the address translation logic for presenting an output address,
the address translation logic operable to receive a first input address from the address input lines and to generate a first output address on the address output lines that is different from the first input address if a portion of the first input address contains a predetermined value.
6. The data processor of claim 5 wherein the address translation logic is further configured to detect a predetermined value present in the high order bits of the input address.
7. The data processor of claim 5 wherein the memory management unit further comprises a translation table, the translation table comprising M-bit values, the translation table addressed by an N-bit value, where M>N, wherein the translation table produces one of the M-bit values based on N lines of the address input lines, wherein the M-bit value produced is used in M lines of the address output lines.
8. The data processor of claim 7 wherein the remaining lines in the address output lines come from the address input lines.
9. A method for transferring data in a data processing unit comprising:
accessing a first value from a first data storage component of the data processing unit;
combining the first value with a base address to produce a first address, the base address representative of an area in a memory;
accessing a second value from a second data storage component of the data processing unit;
accessing a third value from a third data storage component of the data processing unit;
combining the second value with a value stored in a location in the memory indicated by the first address to produce a fourth value;
comparing the third value with the fourth value to produce a comparison result; and
based on the comparison result, either accessing the memory beginning at a location indicated by a second address or performing an exception operation, the second address being computed from the base address.
10. The method of claim 9 further including one or more steps of storing the first value in the first data storage component, storing the second value in the second data storage component, and storing the third value, wherein the one or more steps of storing are performed only at the time the data processing unit is initialized.
11. The method of claim 9 wherein one or more of the first value, the second value, and third value is stored respectively in the first data storage component, the second data storage component, and the third data storage component in a non-volatile manner.
12. The method of claim 9 further comprising loading one or more of the first data storage component, the second data storage component, and the third data storage component respectively with the first value, the second value, and third value.
13. The method of claim 9 wherein the second address is computed based on the base address and an offset value.
14. The method of claim 13 wherein the second address is computed by adding together the base address and the offset value.
15. The method of claim 9 wherein the exception operation comprises accessing content of a fourth data storage component, wherein subsequent processing by the data processing unit continues from instructions beginning at a location in memory computed based on the content of the fourth data storage component.
16. The method of claim 15 wherein the location in memory where subsequent processing continues is computed by adding the content of the fourth data storage component with an offset value.
17. The method of claim 9 wherein the area in memory represents a data object.
18. The method of claim 9 wherein the step of accessing is a step of copying the accessed content to another area in the memory.
19. The method of claim 9 wherein the step of accessing is a step of writing information to the accessed content.
20. The method of claim 9 wherein the first address is produced by adding together the first value and the base address.
21. A data processing unit comprising:
a memory;
instruction decoder logic for producing control signals in response to decoding program instructions;
data processing logic coupled to receive the control signals and configured to perform logic operations in accordance with the control signals;
an instruction register coupled to the memory and to the instruction decoder, the instruction register operable to store an instruction to be decoded;
a program counter operable to store an address of an instruction to be decoded;
a vector address data area for storing a vector address;
a first data area for storing a first value;
a second data area for storing a second value; and
a third data area for storing a third value,
the instruction decoder configured to assert first control signals in response to decoding a first instruction,
the data processing logic, in response to the first control signals, configured to:
combine the first value with a first operand contained in the first instruction to produce a first address;
combine the second value with a value stored in the memory at a location indicated by the first address to produce a fourth value;
compare the third value with the fourth value to produce a comparison result; and
based on the comparison result, either access the memory beginning at a location indicated by a second address or update the program counter with an address based on the vector address, the second address being computed from the first operand.
22. The data processing unit of claim 21 wherein the memory is a re-programmable, non-volatile memory.
23. The data processing unit of claim 21 wherein the memory is one of an EEPROM (electrically erasable programmable read-only memory) and flash memory.
24. The data processing unit of claim 21 wherein the data processing logic, in response to initialization of the data processing unit, is further configured to load data into one or more of the vector address data area, the first data area, the second data area, and the third data area.
25. The data processing unit of claim 21 wherein one or more of the vector address data area, the first data area, the second data area, and the third data area comprise non-volatile memory.
26. The data processing unit of claim 25 wherein the non-volatile memory is one of EEPROM (electrically erasable programmable read-only memory) or PROM (programmable read-only memory).
27. The data processing unit of claim 21 wherein one or more of the vector address data area, the first data area, the second data area, and the third data area comprise hard-wired logic.
28. The data processing unit of claim 21 wherein the instruction decoder is further configured to assert a second set of control signals in response to decoding a second instruction, the data processing logic, in response to the second control signals, is further configured to load data into one or more of the vector address data area, the first data area, the second data area, and the third data area, the data contained in the second instruction.
29. The data processing unit of claim 21 wherein the data processing logic, in response to the first control signals, is further configured to combine the first operand with a second operand contained in the first instruction to produce the second address.
30. The data processing unit of claim 29 wherein the first operand and the second operand are summed to produce the second address.
31. A memory access method in a data processing unit comprising re-programmable non-volatile memory, the method comprising:
detecting a write operation to an area in the re-programmable non-volatile memory;
determining a caching location which identifies an area of memory in a dynamic memory to which data to be written by the write operation can be cached, the area of memory identified by the caching location referred to as a caching area; and
writing the data to the caching area instead of writing the data to the area in the re-programmable non-volatile memory,
wherein the caching area is associated with the area in the re-programmable non-volatile memory.
32. The method of claim 31 wherein the step of determining a caching location comprises:
determining if the data can be stored in the dynamic memory, and if not then:
identifying one or more memory locations in the dynamic memory; and
writing content in the one or more memory locations to the re-programmable non-volatile memory; and
providing an address in the dynamic memory as the caching location in the area of memory.
33. The method of claim 32 wherein the step of determining a caching location in the dynamic memory is limited to determining a caching location in a segment of the dynamic memory.
34. The method of claim 31 further including detecting a second write operation to the area in the re-programmable non-volatile memory and writing the data associated with the second write operation to the area of memory identified by the caching location instead of writing the data to the area in the re-programmable non-volatile memory.
35. The method of claim 31 further including detecting a second write operation to a second area in the re-programmable non-volatile memory, determining if there is a caching area in the dynamic memory that is associated with the second area and if there is then writing data associated with the second write operation to the caching area associated with the second area.
36. The method of claim 31 wherein an address space of the re-programmable non-volatile memory is equal to or greater than an address space of the dynamic memory.
37. The method of claim 32 wherein the step of identifying one or more memory locations is based on the size of the data.
38. The method of claim 32 wherein the step of identifying one or more memory locations is based on contents of the data.
39. The method of claim 38 wherein the step of identifying one or more memory locations includes applying a hash function on the data to produce a hash result, the one or more locations determined based on the hash result.
40. The method of claim 31 wherein the re-programmable non-volatile memory is one of an EEPROM (electrically erasable programmable read-only memory) and a flash memory.
41. The method of claim 31 further including obtaining an address range which defines the area of memory in the dynamic memory.
42. The method of claim 31 further including detecting a read operation from the re-programmable non-volatile memory and reading the re-programmable non-volatile memory to effect the read operation.
43. The method of claim 31 further including detecting a read operation from the re-programmable non-volatile memory and reading one or more memory locations from the area of memory in the dynamic memory to effect the read operation.
44. A data processing unit comprising:
a bus for accessing a dynamic memory;
a re-programmable non-volatile memory;
memory access logic operatively coupled to the memory bus and to the re-programmable non-volatile memory; and
processor logic operatively coupled to the memory access logic to transfer data with the dynamic memory via the memory bus and to transfer data with the re-programmable non-volatile memory,
the memory access logic configured to detect a write operation to the re-programmable non-volatile memory,
the processor logic configured to respond to the memory access logic detecting the write operation and to:
identify a caching location in the area of memory wherein data to be written by the write operation can be cached; and
store the data in the area of memory identified by the caching location.
45. The data processing unit of claim 44 wherein the memory access logic is further configured to:
determine if the data can be stored in the area of memory; and
provide an address in the area of memory as the caching location in the area of memory,
wherein if it is determined that the data cannot be stored in the area of memory, then:
identify one or more memory locations in the area of memory; and
write content in the one or more memory locations to the re-programmable non-volatile memory,
wherein the one or more memory locations are available for caching a write operation.
46. The data processing unit of claim 44 wherein an address space of the re-programmable non-volatile memory is equal to or greater than an address space of the dynamic memory.
47. The data processing unit of claim 44 wherein the re-programmable non-volatile memory is one of an EEPROM (electrically erasable programmable read-only memory) and a flash memory.
48. The data processing unit of claim 44 wherein the dynamic memory is a random access memory.
49. The data processing unit of claim 44 wherein the memory access logic is further configured to detect a read operation from the re-programmable non-volatile memory and to effect a read operation therefrom.
50. The data processing unit of claim 44 wherein the memory access logic is further configured to detect a read operation from the re-programmable non-volatile memory and to access the area in memory in the dynamic memory to effect the read operation.
51. A memory access method comprising:
detecting a write operation to a re-programmable non-volatile memory; and
if a destination address associated with the write operation is within a first range of addresses, then performing a fast write operation of data associated with the write operation to the re-programmable non-volatile memory.
52. The method of claim 51 wherein the first range of addresses spans the entire address space of the re-programmable non-volatile memory.
53. The method of claim 51 wherein if the destination address is not within the first range of addresses, then performing a slow write operation of the data to the re-programmable non-volatile memory.
54. The method of claim 51 wherein the first range of addresses spans a range of addresses less than the address space of the re-programmable non-volatile memory.
55. The method of claim 51 wherein the step of performing a fast write operation is performed if the destination address falls within any of a plurality of ranges of addresses.
56. A memory access method comprising:
detecting a write operation to a non-volatile memory;
determining an access mode;
if the access mode is a first mode, then performing a fast write operation of data associated with the write operation to the non-volatile memory;
if the access mode is a second mode, then performing a slow write operation of the data associated with the write operation; and
if the access mode is a third mode, then:
if the destination address associated with the write operation is within a first range of addresses, then performing a fast write operation of the data associated with the write operation to the non-volatile memory;
if the destination address is not within the first range of addresses, then performing a slow write operation of the data to the non-volatile memory; and
if the access mode is a fourth mode, then:
determining a caching location which identifies an area of memory in a dynamic memory to which data to be written by the write operation can be cached, the area of memory identified by the caching location referred to as a caching area; and
writing the data to the caching area instead of writing the data to the area in the non-volatile memory,
wherein the caching area is associated with the area in the non-volatile memory.
57. A memory control logic for accessing a re-programmable non-volatile memory comprising:
write detection logic to determine when a write operation to the memory is being performed;
address detection logic configured to indicate a detected address when a destination address associated with a write operation falls within a first range of addresses; and
write-control logic operatively coupled to the address detection logic and configured to perform a shallow-write operation of data associated with the write operation when the address detection logic indicates a detected address.
58. The memory control logic of claim 57 further comprising a first address register and a second address register, wherein the first range of addresses comprises data stored in the first and second address registers.
59. The memory control logic of claim 57 further comprising a plurality of register pairs, each register pair representative of a range of addresses in an address space of the memory, wherein the address detection logic is further configured to indicate a detected address when the destination address falls within any of the ranges of addresses represented by the plurality of register pairs.
60. The memory control logic of claim 57 wherein the re-programmable non-volatile memory is one of an EEPROM (electrically erasable programmable read-only memory) and a flash memory.
61. A memory control logic for accessing a re-programmable non-volatile memory comprising write detection logic to determine when a write operation to the memory is being performed and write-control logic configured to perform a shallow-write operation of data associated with the write operation.
62. The memory control logic of claim 61 wherein the re-programmable non-volatile memory is one of an EEPROM (electrically erasable programmable read-only memory) and a flash memory.
63. A memory access method for accessing a re-programmable non-volatile memory comprising:
detecting a write operation to the re-programmable non-volatile memory, the write operation having an associated destination address and associated one or more data to be written to the re-programmable non-volatile memory;
determining an operating mode;
if the operating mode indicates a first mode of operation, then:
performing a partial write operation of each datum to the re-programmable non-volatile memory, if the destination address is within a first range of addresses; and
performing a full write operation of each datum to the re-programmable non-volatile memory, otherwise;
if the operating mode indicates a second mode of operation, then performing a full write operation of each datum to the re-programmable non-volatile memory;
if the operating mode indicates a third mode of operation, then performing a partial write operation of each datum to the re-programmable non-volatile memory; and
if the operating mode indicates a fourth mode of operation, then:
determining a caching location which identifies an area of memory in a dynamic memory where the data can be cached; and
writing the data in the area of memory identified by the caching location instead of writing to the re-programmable non-volatile memory.
64. The method of claim 63 wherein the re-programmable non-volatile memory is one of an EEPROM (electrically erasable programmable read-only memory) and a flash memory.
Description
    BACKGROUND OF THE INVENTION
  • [0001]
    The present invention is related generally to computer processors and more specifically to computer processor enhancements suitable for virtual machines.
  • [0002]
    The notion of a virtual machine has been a topic of academic research for many years. However, the idea has been popularized by Sun Microsystems, Inc. with its Java programming language and runtime environment. The Java virtual machine is software that acts as an interface between compiled Java binary code (known as bytecode) and the underlying hardware platform (e.g., a microprocessor) that actually implements the bytecodes. Once a Java virtual machine has been provided for a platform, any Java program can run on that platform.
  • [0003]
    As the name implies, a virtual machine is an abstraction (a virtual representation) of a physical processor and typically comprises virtual counterparts of a conventional processor. For example, the Java virtual machine specifies elements such as an instruction set, a set of registers, a stack, a “garbage heap,” and a method area. The real implementation of this abstract or logically defined processor is the native machine code executed by the real processor or be built into the microchip processor itself (so-called microcode).
  • [0004]
    An area in which the virtual machine is finding application is in the use of smartcards. Typically, a smartcard is a card that is embedded with either a microprocessor and a memory chip or only a memory chip with non-programmable logic. A microprocessor-based smartcard can add, delete, and otherwise manipulate information on the card, and is readily adaptable to many functions. For example, the national health card of Taiwan is a general-purpose smartcard which can store multiple programs downloaded from multiple vendors. Other examples include a credit card which can hold multiple “loyalty” applications to reward the user for multiple purchases from a single vendor, and which can also function as an employee ID card. Java Card is based on the underlying technology of the Java programming language. The Java virtual machine specification and the language definition have been accommodated to make Java technology available in a smartcard.
  • [0005]
    The present invention is motivated in part by the potential benefits of virtual machine based smartcards and, in general, by the use of virtual machine technology in resource-limited environments such as the smartcard.
  • [0006]
    For example, a smartcard typically uses three kinds of memory: ROM (read-only memory), RAM (random-access memory), and EEPROM (electrically erasable programmable read-only memory). A programming language (e.g., Java) used in a virtual machine is an abstract language and typically uses a memory model that does not recognize different kinds of memory. The translation from the abstract memory space of the virtual machine environment to the physical memory contained in or connected to the smartcard is performed by the virtual machine.
  • [0007]
    Java is an object oriented language, and thus relies heavily on the creation and destruction of objects. In an object oriented language, there are no facilities for supporting procedures (i.e., functions without an associated object), so it is not possible to do procedural programming. Therefore, it is necessary to create, destroy, and call objects to perform a given task.
  • [0008]
    Typically, smartcards have very limited RAM capacity; 2K (bytes) to 4K is common. If the programming objects are created in the limited RAM, then it becomes very difficult to perform complex tasks because the amount of memory available for object creation is very limited. A smartcard typically has much more EEPROM capacity than RAM, usually on the order of 32K to 96K of EEPROM as compared to the 2K-4K of RAM. Consequently, the virtual machine uses EEPROM for instantiation of object instance data, and uses RAM only for the Java stack and “transient RAM arrays.”
  • [0009]
    The allocation of EEPROM memory for object instance data permits the creation of large programs because there is much more EEPROM capacity than RAM capacity. However, there is a substantial cost for this extra capacity. RAM typically requires one clock cycle to read or write data. EEPROM reads, likewise, typically are performed in one clock. EEPROM write operations, on the other hand, typically require about 4 mS to complete, or about 14,289 clock cycles running at 3.57 MHz! Object creation involves initialization of the data comprising the object, which require writes to the EEPROM. It can be appreciated therefore that object creation and object instance data updates is a very, very slow process.
  • [0010]
    Similarly, the virtual machine heap is allocated from EEPROM. The heap is a dynamic data structure which is constantly read and, more significantly, written. Thus, writing to dynamically allocated memory is a very slow process. Generally, any variable not on the stack will be allocated from the EEPROM heap, and so writes to non-stack variables will be very slow.
  • [0011]
    Another area of concern is in the area of providing adequate memory management. In the context of a smartcard environment, and other similarly constrained operating environments, system memory capacity is a very limited resource. Historically, limited memory capacity has always been a problem in processors, especially on multitasking systems where multiple users can run different programs simultaneously as if each user were the only person on the system. The earliest multitasking systems supported a method of multitasking memory called “swapping.” This system saves the entire executing state of a low priority program to a disk, thus freeing up memory to allow another, higher priority program to execute. Swapping can also take place when a program is idle, waiting for data transfer with a slow I/O device (e.g., tape drive) to complete.
  • [0012]
    Though swapping is effective in providing access to memory, the performance is less than optimal because the entire address space of the program must be swapped to and from the disk. The idea of swapping out entire programs evolved into the notion of “paging” where the address space of a program was divided into some number of fixed-size pages. Each of these pages could then be swapped out to disk and back, as required by the system, without having to swap out the entire program thus improving system performance.
  • [0013]
    Memory protection mechanism are needed when multiple programs are executing on the processor. Memory protection via hardware instructions has traditionally been implemented in one or two different ways: The first method was via extra data bits added to each machine word. This method required special hardware and was incompatible with industry standards, and was soon discontinued in favor of memory management logic. This method relies upon a component called a memory management unit (MMU) to implement memory protection. The MMU approach provides page-level granularity, usually on the order of 4 Kbytes and higher.
  • [0014]
    Another memory management technique concerns memory access speed. The “cache” is a high-speed memory and was originally envisioned as a slave to a slower core memory in such a way that in practical cases the effective access time is nearer that of the fast memory than that of the slow memory. Caches have traditionally been used to decrease the average access time of slower RAM memory, and have also been used in mechanical storage media such as disk drives.
  • [0015]
    There is a need to provide a suitable operating environment for operating a virtual machine in a smartcard environment and in general in any similar resource-limited environment other than in a smartcard application.
  • SUMMARY OF THE INVENTION
  • [0016]
    The present invention provides a specialized kind of instruction for improving the access time to data structures. In one aspect of the invention a security feature is provided. In another aspect of the invention, access time improvements are provided to accommodate the security features.
  • [0017]
    The present invention further provides a data processing architecture featuring memory access techniques for accessing re-programmable, non-volatile memory. In one aspect of the invention, a caching mechanism for accessing such memories is provided. In another aspect of the invention, a mechanism for writing to such memories is provided.
  • [0018]
    The present invention still further provides a memory management technique that can be adapted to enhance virtual machine implementations.
  • BRIEF DESCRIPTION OF THE DRAWINGS
  • [0019]
    The present invention can be appreciated by the description which follows in conjunction with the following figures, wherein:
  • [0020]
    FIG. 1 shows a generalized block diagram of a system incorporating data processing aspects of the present invention;
  • [0021]
    FIG. 2 represents an illustrative embodiment of a data processing unit in accordance with the present invention;
  • [0022]
    FIG. 3 shows an illustrative embodiment of memory access in accordance with the present invention;
  • [0023]
    FIG. 3A illustrate a memory mapping according to an aspect of the invention as embodied in FIG. 3;
  • [0024]
    FIG. 4 shows a memory map exemplar according to an aspect of the invention as embodied in FIG. 3;
  • [0025]
    FIG. 4A is a graphic illustration of the mapping process;
  • [0026]
    FIGS. 5, 5A, and 6 show an embodiment of a read access operation in accordance with the present invention;
  • [0027]
    FIGS. 7 and 7A show an embodiment of a write access operation in accordance with the present invention;
  • [0028]
    FIGS. 8-13 show an illustrative embodiment of a caching mechanism in accordance with the present invention;
  • [0029]
    FIGS. 14 and 15 illustrate an embodiment of a write operation in accordance with the present invention;
  • [0030]
    FIGS. 16 and 17 show additional aspects of the write operation of the present invention; and
  • [0031]
    FIG. 18 is a schematic representation of a typical re-programmable, non-volatile memory cell.
  • DESCRIPTION OF THE SPECIFIC EMBODIMENTS
  • [0032]
    FIG. 1 depicts an illustrative embodiment of a data processing unit 10 according to the present invention incorporated in a high-level generalized block diagram representation of a computing environment. Various aspects of the invention as embodied in the data processing unit 10 are shown in the following figures. A bus 14 is shown connecting together the data processing unit and other components. The bus 14 is a representation of the various physical connections that may be needed among the components, including but certainly not limited to a data bus, signal lines for asserting control signals among the components, power lines for providing power as needed to the components, and so on.
  • [0033]
    The components shown in FIG. 1 are typical but not exhaustive of elements in a computing environment. A read-only memory (ROM) 11 can be provided. The ROM can contain program code and data as required in a given system. A random access memory (RAM) 12 can be provided. Dynamic RAMs are commonly used for providing a transient area of storage during operation. One or more peripherals 13 can be provided as well.
  • [0034]
    FIG. 2 is a generalized block diagram representation of an illustrative embodiment of the data processing unit 10 in accordance with the invention. The data processing unit comprises a processor logic component configured to operate in accordance with the invention. An illustrative embodiment of the processor logic can comprise an arithmetic and logic unit (ALU) 24 and control logic 23. The ALU 24 typically comprises logic configured to implement the various operations specified by the instruction set of the data processing unit. The control logic 23 typically comprises logic configured to provide various control signals to coordinate the operation of the ALU and other components to execute the instructions comprising a program. It can be appreciated the control signals are distributed by way of control signal lines, schematically represented in the figure generally as the various lines connecting the components. One such control signal is an interrupt signal 24 a from the ALU provided to the control logic.
  • [0035]
    An instruction decoder component 22 comprises logic configured to receive and decode the instruction set of the data processing unit 10. The instruction decoder component produces control signals and data signals depending on the particular instruction being decoded and passes the signals, via the signal lines shown schematically in the figure, to the control logic 23 which then produces appropriate control signals to effect execution of the instruction.
  • [0036]
    The data processing unit 10 further comprises a plurality of data storage components. In a particular embodiment, the data storage components are data registers 21 a-21 h. For example, an instruction register 21 a can be provided to store an instruction to be fed to the instruction decoder 22. A set of general purpose registers 21 b can be provided for storing data. Typically, the instruction set of the data processing unit includes data manipulation instructions such as ADD, SUBTRACT, COMPARE, and the like. These instructions can use the general purpose registers 21 b to store and retrieve data. A program counter 21 c is a register that contains the address of the next instruction to be fetched into the instruction register 21 a. FIG. 1 shows additional data storage components, namely, a PROTINDEX register 21 d, a PROTMASK register 21 e, a PROTID register 21 f, a vector-based register (VBR) 21 g, and an EEPROM_MODE register 21 h. The purposes of these registers will be discussed below in the appropriate portions of the specification.
  • [0037]
    The data processing unit 10 further comprises a re-programmable, non-volatile memory. The particular embodiment illustrated in FIG. 2 shows an EEPROM memory 26 as the re-programmable, non-volatile memory. Alternative configurations can include other similar re-programmable, non-volatile memories such as flash memory. A memory controller is embodied as an EEPROM controller 27 which is configured to provide access to the EEPROM memory 26 according to the invention. It is understood that suitable control lines and internal data buses are present to interconnect the various components. Such lines are schematically represented in FIG. 2.
  • [0038]
    In addition to the EEPROM memory 26, access to an external memory (e.g., dynamic RAM) can be provided via a suitable memory bus 14 a. A memory management unit (MMU) 25 constitutes an illustrative embodiment of a memory access logic component of the data processing unit 10. Memory accesses required by the ALU 24 can be handled by the MMU 25. The MMU is configured to perform address translation in accordance with the present invention. As can be seen in the embodiment shown in FIG. 2, the MMU also serves to provide access (via the EEPROM controller 27) to the EEPROM memory 26.
  • [0039]
    FIG. 3 is a schematic representation of an illustrative embodiment of the MMU 25 in accordance with the invention. As indicated in the figure, a 24-bit addressing scheme is used in a particular embodiment of the data processing unit 10. It can be appreciated of course that other address sizes can be easily accommodated in accordance with the invention.
  • [0040]
    FIG. 3 shows that address signals from the ALU 24 are provided to the MMU 25. In a particular embodiment, this can be accomplished by providing address lines from the ALU to the MMU. The MMU provides address signals to the memory bus 14 a shown in FIG. 2 which in a particular embodiment can comprise MMU address lines 34 a-34 c being coupled to the memory bus.
  • [0041]
    The MMU 25 comprises a data store 31 to provide a value. In the embodiment shown, a value of 0B0 is provided. The value from the data store 31 is provided to a comparator 30. The comparator performs a comparison of the value from the data store 31 against a subset of the address bits. In the embodiment illustrated, the uppermost 8-bits of the 24-bit address (A23-A16) are compared by the comparator 30. An output signal 30 a of the comparator is used to control a data switch 32 which, in a particular embodiment of the invention, can be a suitably configured multiplexer.
  • [0042]
    The MMU 25 further comprises a data translation table component 33. The data table is configured to produce an output value 33 b based on an input value 33 a. In a particular embodiment of the invention, another subset of the address bits serves as the input to the data table 33, namely, address bits A15-A8. The output 33 b of the data table is a 16-bit value provided over 16 bitlines. The 16 bitlines can be divided into upper 8 bits and lower 8 bits.
  • [0043]
    Address lines representing address bits A23-A16 are provided to an input 32 a of the data switch 32. The upper 8 bits from the data table component 33 are provided to an input 32 b of the data switch. The data switch is configured to provide the input 32 a to its output 32 c when the comparator output signal 30 a is in a first data state. The data switch is configured to provide the input 32 b to the output 32 c when the comparator output signal is in a second data state.
  • [0044]
    Address lines representing address bits A15-A8 are provided to an input 32 d of the data switch 32. The lower 8 bits from the data table component 33 are provided to an input 32 e of the data switch. The data switch is configured to provide the input 32 d to its output 32 f when the comparator output signal 30 a is in a first data state. The data switch is configured to provide the input 32 e to the output 32 f when the comparator output signal is in a second data state.
  • [0045]
    The MMU 25 is configured to pass still a third subset of address bits directly to the MMU address lines 34 a, namely, address bits A7-A0.
  • [0046]
    FIG. 3A illustrates the operation when there is no address translation. When the uppermost 8 bits of the address (A23-A16) from the ALU 25 does not contain the value 0B0, the comparator 30 is configured to place the comparator output signal 30 a in the first data state. Consequently, the data switch 32 simply passes the address lines A23-A8 from the ALU 25 directly to the MMU address lines 34 c and 34 b. The address lines A7-A0 are always passed directly to the MMU address lines 34 a. Consequently, no address translation occurs. This is shown in FIG. 3A by the dashed lines.
  • [0047]
    FIG. 3B illustrates the translation operation of the MMU 25. When the uppermost 8 bits of the address (A23-A16) from the ALU 25 does contain the value 0B0, the comparator 30 is configured to place the comparator output signal 30 a in the second data state. Consequently, the data switch 32 passes the output lines 15-8 from the data table 33 to the MMU address lines 34 c. Similarly, the output lines 7-0 from the data table are passed to the MMU address lines 34 b. The address lines A7-A0 again are passed directly to the MMU address lines 34 a.
  • [0048]
    FIG. 4 shows a memory map exemplar illustrating a mapping from non-contiguous physical memory to a continuous memory space that can be provided by the MMU 25. FIG. 4 also shows how this mapping scheme can be applied to provide the address space for a virtual machine. First, a discussion of the mapping. The figure shows a memory map 42 of physical memory. The physical memory comprises three kinds of memory: an EEPROM, a RAM, and a ROM. The address space of the physical memory, which in this particular implementation is 24 bits, is partitioned among the three kinds of memory. In the exemplar shown in the FIG. 4, the ROM occupies the low order address space and the EEPROM occupies the high order address space, allocating RAM memory to the middle portion of the address space. A 64K fragment 41 of the physical memory is shown enlarged. The address space of the memory fragment extends from 0B0FFFF to 0B00000.
  • [0049]
    Consider the MMU detail shown in FIG. 3. When an address (say, 08760) is provided from the ALU 24 to the MMU 25, it can be seen that the MMU will simply pass this address to the bus 14 a (FIG. 2) unchanged, to access the content of memory location 08760 for a write operation or a read operation. Incidentally, for the particular example shown, memory location 08760 is in ROM, so the access is likely to be a read operation.
  • [0050]
    Now, consider when the ALU 24 provides an address, say 0B0b3b2b1b0, where bn, is a 4-bit value. The MMU 25 will map that address to another address having the form 0N3N2N1N0b1b0, where Nn, is a 4-bit value. As can be seen in FIG. 3, the low order 8 bits (0b1b0) are passed through. However, the data table 33 will provide a new 16-bit value (0N3N2N1N0) to replace the original 16-bit value (0B0b3b2) that came from the ALU. Moreover, the new 16-bit value is selected based on the middle 8 bits (0b3b2). Thus, the MMU can map each of the 256 contiguous 256-byte pages comprising the 64K memory fragment 4, to any 256-byte page in the physical memory 42 falling on a two-byte boundary. For example, the 256-byte page 43 a is shown mapped to a 256-byte page in physical memory, which can be any page in the physical memory address space subject to the page boundary constraint. The same holds true for the other 256-byte pages comprising the fragment 41. The segment 43 f shows that a contiguous collection of two or more 256-byte pages can mapped to another area of physical memory of the same size, highlighting the idea that it is not necessary to scatter each 256-byte page all over the physical address space.
  • [0051]
    If the data store 31 which supplies the constant value 0B0 is changed, then a different 64K fragment can be mapped. It is noted that a 256-byte page in the fragment 41 can be mapped to itself, or to another 256-byte page within the fragment 41. It can be seen that the contents of the data table (translation table) 33 determine the mapping function.
  • [0052]
    Thus, memory control logic embodied in accordance with the present invention, such as MMU 25, has the interesting ability to partition a contiguous portion of its physical memory address space into plural segments and to map each such segment to an equal-sized segment anywhere in its physical memory. In other words, the MMU can provide address translation for only a subset of the address space. It can be seen that the 256-byte page size granularity can be varied to be any page size, from one byte granularity to something larger (e.g., another typical page size might be 128 bytes). Referring to FIG. 3, the page size can be varied in the MMU by regrouping the address lines A23-A0 (from the ALU).
  • [0053]
    Merely as an illustration, consider an extreme (and unlikely) example where the MMU is reconfigured in the following manner (referring to FIG. 3):
      • the G2 address lines consist of A15 and A2 instead of A15-A8 and
      • the G3 address lines consist of A1 and A0 instead of A7-A0 (4-byte pages).
        The data table 33 would provide 4048 (214) 22-bit values, allowing the MMU to map each of 4048 4-byte “pages” contained in a 64K block to any other 4-byte page in the physical memory, with the constraint that the map 4-byte page falls on a 4-byte boundary.
  • [0056]
    In accordance with another implementation, the number of G1 address lines can be increased or decreased in number to vary the possible locations of the fragment 41. As shown in FIG. 3, the G1 address lines comprise the upper 8 bits of a 24 bit address space. As shown, the MMU 25 of FIG. 3 limits the fragment 41 to be on a 64K boundary. It can be appreciated, however, that numerous other permutations can be made to the address lines provided to the MMU to achieve various effects.
  • [0057]
    Consider once again the memory map of FIG. 4, this time with the notion of a virtual machine in mind. The figure shows that a virtual machine monitor component VM is loaded into the memory, and in particular in the RAM memory. The VM occupies the physical address space from 0B0FFFF to (0B0h3h2h1h0+1). A virtual machine RAM memory (VM RAM) occupies the address space from 0B0h3h2h1h0 to 0B00000. This configuration gives the virtual machine a virtual address space of 64K which occupies the physical address space from 0B0FFFF to 0B00000. The memory fragment 41 therefore represents the virtual address space of the virtual machine. The address range shown in the far left portion of FIG. 4 represents the virtual address range 0FFFF to 00000 of the virtual machine.
  • [0058]
    When the virtual machine is initialized, one of the initialization routines performs an initialization of a translation table. Thus, in an embodiment of this aspect of the present invention, the translation table is a data storage component (e.g., data translation table 33) in the data processing unit 10 that has an associated address within the physical address space of the data processing unit. The MMU 25 is configured to recognize this address and perform the requested I/O with the table. Thus, the following code fragment exemplar can be used to initialize the table as part of the initialization sequence of the virtual machine:
    void VM_init( )
    {
      for( i = ROM_START; i < ROM_END; ++i )
        translate[i] = &ROM[i];
      for( i = RAM_START; i < RAM_END; ++i )
        translate[i] = &RAM[i];
    }

    where ROM[ ] and RAM[ ] are arrays which hold the mapping of each virtual page to a physical page, and translate[ ] is the translate table 33.
  • [0059]
    Operation of the virtual machine amounts to execution, by the data processing unit 10, of the code which implements the virtual machine monitor stored in the physical memory 42 between 0B0FFFF and (0B0h3h2h1h0+1). An applet to be executed by the virtual machine is loaded into the VM RAM area. When the data processing unit executes the virtual machine code, address references are addresses in the physical memory address space 42 of 0FFFFFF−00. However, when the virtual machine “executes” the applet code, address references are addresses in the virtual memory address space of 0FFFF−00. The virtual machine code must perform a translation of the virtual address into a real address.
  • [0060]
    The virtual machine includes functionality to perform translations of a virtual address to a physical address. The following code fragment can be used:
    char VM_read_byte(unsigned int addr)
    {
      char *p = (char *)((addr & 0xff)|(translate[addr >> 8]));
      return (*p);
    }

    This subroutine is executed in the physical address space and will return a value representing an address in the physical address space 42. Since the VM executes in a 24-bit address space where the addresses begin with 0B0, The MMU 25 will resolve the address reference:
      • ((addr & 0ff)|(translate[addr >>8]))
        in accordance with the mapping discussed above. FIG. 4A illustrates this mapping process graphically.
  • [0062]
    FIGS. 5, 5A, and 6 illustrate a particular embodiment for reading data in accordance with another aspect of the present invention. FIG. 5 highlights the control sequence for effecting a protected read operation according to the invention. FIG. 6 highlights (with heavy lines) the components of the data processing unit 10 shown in FIG. 2 that participate in the read operation. FIG. 5A illustrates schematically the operation of the PMOV instruction on an object.
  • [0063]
    An instruction exemplar in accordance with this aspect of the invention can be represented by the following assembly code formatted instruction:
      • PMOV .L @(K, Rm), Rn
        where:
      • PMOV is a mnemonic for the data transfer instruction;
      • .L is indicative of the size (or number) of data elements (e.g., bytes) to be transferred;
      • (K, Rm) represents a source address formed by adding K to the contents of register Rm;
      • Rm is a pointer to the object (FIG. 5A);
      • Rn is the destination register; and
      • the @ sign is conventional notation for indirect addressing.
        It can be appreciated of course, that this instruction exemplar is merely illustrative and is useful for explaining this aspect of the invention. However, one of ordinary skill in the relevant art will readily appreciate that there are myriad possible variations of the instruction. For example, the data size can be specified differently. The amount of data involved in the transfer can be any appropriate size; e.g., the “.L” qualifier typically specifies the number of bytes to be transferred. Other qualifiers such as “.B”, “.W”, and so on can be provided. More generally, a PMOV instruction can be provided which performs block-level transfers (e.g., 256 bytes), if desired. Alternate addressing modes can be provided. Here, indirect addressing is provided only for the source; however, the destination can also be indirectly addressed. As implied, the first operand is the source operand and the second operand specifies a destination. Of course, the operands can be reversed.
  • [0071]
    When the instruction decoder 22 decodes the above PMOV instruction, signals corresponding to the decoded instruction are provided to the control logic 23 to effect the read operation. In a particular embodiment of this aspect of the invention, the control logic 23 is configured to assert control signals to perform the sequence of actions shown in FIG. 5. First, in a step 51, the ALU 24 is controlled to fetch the value contained in the register Rm. This is the address of the object from which the datum will be read. The value contained in the data storage component PROTINDEX 21 d is added to the address from Rm to obtain a second address. The second address is provided to the MMU 25, where it may be translated into another address. The address provided by the MMU is then used to retrieve the content contained in the addressed memory location. Recall from the discussion above, that the addressed memory location can be anywhere in the memory space of the data processing device. In the particular embodiment of this aspect of the invention, the content retrieved is an 8-bit value which is provided to the ALU 24.
  • [0072]
    In a step 52, the ALU 24 is then controlled to fetch the value contained in the data storage component PROTMASK 21 e and to logically AND the value with the retrieved content. The result is compared to the value contained in PROTID 21 f in a step 53.
  • [0073]
    If a match is found, the ALU 24 is then controlled, in a step 55, to add the value K provided by the instruction decoder 22 with the address contained in the source register Rm, to obtain yet a third address where the data to be read is located. The datum is accessed and stored in the destination register Rn as indicated in FIG. 5A.
  • [0074]
    If the comparison from step 53 does not result in a match, then the ALU 24 is controlled, in a step 54, to trigger an exception by asserting the control signal 24 a. The control signal is provided to the control logic 23. According to a particular implementation of the invention, the control logic is configured to assert further control signals in response to the control signal 24 a in order to control the ALU 24 retrieve the value contained in the data storage component VBR 21 g and to add the value with a constant value supplied by the control logic to produce an address. The address is then loaded into the program counter 21 c and execution resumes at that address.
  • [0075]
    It can be appreciated from the foregoing, that the “protected” aspect of this read operation arises from the operations involving the PROTMASK 21 e and the PROTID 21 f. One of ordinary skill will readily appreciate that the processing performed in step 54 amounts to an operation sometimes referred to as a vectored interrupt. The VBR register 21 g represents a pointer to a table of addresses (interrupt vector table). It is understood and appreciated by those of ordinary skill in the art that the vector table contents are typically defined by the application, whether dynamically or at boot up. In the case of a read violation with the PMOV command, the constant value provided by the control logic 23 represents a particular entry in the vector table associated with this particular read violation. It can be appreciated that this constant value can be programmable, and initialized at boot up or dynamically updated (e.g., from memory) instead of being hard-coded in the control logic 23.
  • [0076]
    FIGS. 7, 7A, and 6 illustrate a particular embodiment for writing data in accordance with the present invention. FIG. 7 highlights the control sequence for effecting a protected write operation according to the invention. FIG. 6 highlights (with heavy lines) the components of the data processing unit 10 shown in FIG. 2 that participate in the write operation. FIG. 7A illustrates schematically the operation of the PMOV instruction on an object.
  • [0077]
    An instruction exemplar in accordance with this aspect of the invention can be represented by the following assembly code formatted instruction:
      • PMOV .L Rn, @(K, Rm)
        where:
      • PMOV is a mnemonic for the data transfer instruction;
      • .L is indicative of the size (or number) of data elements (e.g., bytes) to be transferred;
      • Rm points to the object to be written to (FIG. 7A);
      • (K, Rn) K represents a location in the object and register Rn is a source; and
      • the @ sign is conventional notation for indirect addressing.
        It is noted here, as in the case of the read operation above, that this instruction exemplar is merely illustrative and is useful for explaining this aspect of the invention. However, one of ordinary skill in the relevant art will readily appreciate that there are myriad possible variations of the instruction. For example, the data size can be specified differently. The amount of data involved in the transfer can be any appropriate size; e.g., the “.L” qualifier typically specifies the number of bytes to be transferred. Other qualifiers such as “.B”, “.W”, and so on can be provided. More generally, a PMOV instruction can be provided which performs block-level transfers (e.g., 256 bytes), if desired.
  • [0084]
    In this particular embodiment of this aspect of the invention, the same mnemonic, PMOV, is used for read operations and write operations. Therefore, in order to distinguish between the two operations, the syntax for specifying the operands is different between read operations and write operations. This allows an assembler to make the distinction and to produce the proper opcodes to represent either a read operation or a write operation. It can be appreciated that separate mnemonics can be used, one for read and another for write. Alternatively, a different syntax convention can be used.
  • [0085]
    When the instruction decoder 22 decodes a PMOV instruction, signals corresponding to the decoded instruction are provided to the control logic 23 to effect the write operation. In a particular embodiment of this aspect of the invention, the control logic 23 is configured to assert control signals to perform the sequence of actions shown in FIG. 7. First, in a step 71, the ALU 24 is controlled to fetch the value contained in the register Rm (i.e., a pointer to the object) and the value contained in the data storage component PROTINDEX 21 d, and to add the two values to obtain a second address. The second address is provided to the MMU 25, where it may be translated into another address. The address provided by the MMU is then used to retrieve the content contained in the addressed memory location. In the particular embodiment of this aspect of the invention, the content retrieved is an 8-bit value which is provided to the ALU 24.
  • [0086]
    In a step 72, the ALU 24 is then controlled to fetch the value contained in the data storage component PROTMASK 21 e and to logically AND the value with the retrieved content. The result is compared to the value contained in PROTID 21 f in a step 73.
  • [0087]
    If a match is found, the ALU 24 is then controlled, in a step 75, to add the value K provided by the instruction decoder 22 with the content of the register Rm, to obtain a computed address indicating the area in memory (i.e., in the object) to which data associated with the write operation will be written. The control logic 23 then asserts control signals to perform a write operation of data contained in the source register Rn, to the area in memory indicated by the computed address, as illustrated in FIG. 7A. It can be appreciated that the source register Rn can contain the data (e.g., comprising one or more bytes) to be written, or it can point to an location in memory of the source data. Alternatively the source register can represent a beginning address in memory of a block of data to be written into the object.
  • [0088]
    If the comparison from step 73 does not result in a match, then the ALU 24 is controlled, in a step 74, to trigger an exception by asserting the control signal 24 a. The control signal is provided to the control logic 23. According to a particular implementation of the invention, the control logic is configured to assert further control signals in response to the control signal 24 a in order to control the ALU 24 retrieve the value contained in the data storage component VBR 21 g and to add the value with a constant value supplied by the control logic to produce an address. The address is then loaded into the program counter 21 c and execution resumes at that address.
  • [0089]
    It can be appreciated from the foregoing, that the “protected” aspect of the write operation arises from the operations involving the PROTMASK 21 e and the PROTID 21 f. One of ordinary skill will readily appreciate that the processing performed in step 74 amounts to an operation sometimes referred to as a vectored interrupt. As discussed above, the VBR register 21 g represents a pointer to a table of addresses (interrupt vector table). In the case of a write violation with the PMOV command, the constant value provided by the control logic 23 represents a particular entry in the vector table associated with this particular write violation. The constant value can be the same as that for the read violation, or not. It can be further appreciated that this constant value can be programmable, and initialized at boot up or dynamically updated (e.g., from memory) instead of being hard-coded in the control logic 23.
  • [0090]
    Referring now to FIGS. 8-13, an illustrative embodiment of a memory access method in accordance with an aspect of invention is described. The data processing unit 10 further comprises an EEPROM_MODE data storage component 21 h. The content of this register indicates an access mode for the EEPROM memory 26. For example, as illustrated in FIG. 8, an EEPROM_MODE value of “3” can be used to indicate that all read and write accesses to the EEPROM memory space will generate an exception and be handled by an exception handler in a manner to be discussed shortly. It can be appreciated that in other embodiments of this aspect of the invention, the EEPROM_MODE register 21 h can be implemented anywhere in the data processing unit 10 that is convenient. For example, the data store can be provided in the EEPROM controller 27.
  • [0091]
    FIG. 9 highlights with heavy lines the components of the data processing unit 10 illustrated in FIG. 2 involved in generating an EEPROM access exception. An instruction fetched in the instruction register 21 a is decoded by the instruction decoder 22. Suitable control signals produced by the decoded instruction are provided to the control logic 23. The ALU 24 is then controlled according to control signals asserted by the control logic and provided to the ALU. If a memory reference is made, an address is determined by the ALU from the operand(s) of the instruction and provided to the MMU 25. In accordance with a particular embodiment of the invention the EEPROM controller 27 is configured to recognize the address space associated with the EEPROM memory. The EEPROM controller is further configured to check the value in the EEPROM_MODE register 21 h when it recognizes an address as being an EEPROM address. If the EEPROM_MODE is “3”, then the EEPROM controller can be configured to assert the control signal 27 a which is provided to the control logic 23. Appropriate exception handling (e.g., such as described in connection with VBR register 21 g), can be performed by the control logic to continue execution in an appropriate exception handler.
  • [0092]
    The illustrative embodiment shown in FIG. 9 indicates that for EEPROM_MODE “3,” the EEPROM controller 27 is configured to trigger the control signal 27 a on an address that falls within the full range of the EEPROM address space. It can be appreciated of course that the controller can be configured to trigger the control signal on subset of addresses in the EEPROM address space. This can include recognizing an address within one or more predefined (or programmably definable) address ranges. It can be understood that suitable data storage components (e.g., registers) can be provided to implement multiple address ranges.
  • [0093]
    FIG. 10 illustrates a caching method in accordance with the present invention. An application 101 is shown performing I/O operations with an EEPROM memory 103, or some other similar re-programmable, non-volatile memory. A caching mechanism 102 provides a caching function of all I/O operations. In accordance with this aspect of the present invention, the caching mechanism caches write operations in addition to read operations. In a particular embodiment of this aspect of the invention, the caching mechanism as shown in the figure is implemented as an exception handler routine and uses RAM as a cache memory.
  • [0094]
    FIG. 11 illustrates a further aspect of the caching method of the present invention. An application 111 performs I/O with an EEPROM memory 113, or some other similar re-programmable, non-volatile memory. In accordance with this aspect of the present invention, the cache mechanism 112 caches (to a “write” cache, not expressly shown) only write operations to the EEPROM memory. Read operations are performed directly from the EEPROM; with the exception that memory locations updated by cached write operations are read based on the content of the “write” cache. In a particular embodiment of this aspect of the invention, the caching mechanism is implemented as an exception handler routine, and the “write” cache is provided by RAM.
  • [0095]
    Referring again for a moment to FIG. 9, it can be appreciated the upon detecting the control signal 27 a, the control logic 23 can be configured, in a particular embodiment of the present invention, to effect the caching operations shown in FIG. 10 or in FIG. 11. In one particular implementation, the data processing unit 10 can include hard-wired logic to perform the caching operations. Alternatively, the control logic 23 in response to the control signal 27 a can perform exception handling and transfer execution to an exception handler routine.
  • [0096]
    FIG. 12 illustrates processing during a “write” cache operation according to an embodiment of this aspect of the invention. As described in FIG. 9, a write operation 123 to the EEPROM memory 26 by an application can be detected. The cache 124 may have to make space or “evict” other cached memory, operation 122, in order to store the data to be cached. In this aspect of the invention, the data is write data associated with a write operation to an EEPROM memory or other similar re-programmable, non-volatile memory. It can be appreciated that the cache 124 shown in the figure is representative of the cache processing and the cache memory. For example, in a particular embodiment, the cache processing can be a software routine access via an exception. The cache memory can be the RAM memory 12 itself. It is worth noting that in a conventional caching scheme, a high speed cache memory is used to improve the read access of the RAM. It is ironic therefore that, in this particular embodiment of this aspect of the invention, the RAM serves the role of a cache memory.
  • [0097]
    Typically, caches are organized into a number of cache lines, where each cache line holds a specific number of consecutive bytes of memory (referred to as the cache line size). Caches typically organize multiple cache lines into groups based on a simple mapping function; e.g., a hashing algorithm such as address bit hashing. Consequently, the data to be cached can only be stored in a cache line falling into the cache line group as determined by the mapping function. If no cache line is available in that group (cache miss), the cache must evict a cache line out of that group. The organization and eviction mechanisms are standard practice in current cache deployment and, in accordance with an aspect of the present invention, are adapted for the “write” cache. For example, evicting a cache line in accordance with an embodiment of the invention includes performing a write operation to the area in the EEPROM memory to effect write the operation. This then frees up a cache line in the cache to store the data to be cached. The question of when to perform a write operation to the EEPROM can be answered using policies conventionally employed for read caches.
  • [0098]
    If data needs to be cached (in memory 124) from the EEPROM, a similar process 123 of finding an available cache line to hold the data is performed. If this location is presently occupied by a valid data item, then the present contents of the cache line must be “evicted” and written to EEPROM so the data will not be lost prior to filling the cache line with new data. If the cache line presently holds no data, then the cache line can be filled with the new data with no prior preparation. So, there is an eviction process which may or may not be necessary prior to filling the cache line, depending on whether it was or was not holding valid data.
  • [0099]
    FIG. 13 illustrates processing during a “read” operation to the EEPROM memory 26. In one aspect of the invention, the read can be satisfied from the cache, operation 132, if it has the most current “written” data. Otherwise, a read from the EEPROM is performed, operation 132. As indicated in FIG. 10, however, another aspect of the invention provides conventional read caching, in addition to “write” caching. With respect to FIG. 13, then, the additional caching of read operations may impose a heavier burden on the cache mechanism. On the other hand, it may improve read operation throughput. Whether the caching function provides for caching read operations is a matter of design choice based on considerations not relevant to the practice of the present invention.
  • [0100]
    FIGS. 14-16 illustrate example embodiments of a method for writing to EEPROM memory 26, or to other similar re-programmable, non-volatile memory. Refer for a moment to FIG. 18 which shows the construction of a typical reprogrammable, non-volatile memory cell (e.g., EEPROM, flash memory). The figure schematically illustrates the process of the floating gate being loaded (programmed) with electrons during programming operation. The figure shows diagrammatically the flow of electrons from the channel region, across the gate oxide, and into the floating gate structure, where the electrons are stored. The also shows the process of electron tunneling to unload (erase) electrons from the floating gate. By a tunneling effect, the electrons tunnel from the floating gate structure into the n+ well of the drain.
  • [0101]
    Referring to FIGS. 14 and 15, an aspect of the invention provides for increased write times (fast write operations, shallow write operations, non-persistent writes) to an EEPROM memory 26 (FIG. 2), or other similar re-programmable, non-volatile memory.
  • [0102]
    Recall from FIG. 3 that the EEPROM_MODE register 21 h indicates an access mode for the EEPROM memory. Thus, as an embodiment of this aspect of the present invention, an EEPROM_MODE value (say, “2”) can be used to indicate that all write operations to the EEPROM memory are performed using “fast writes.” Thus, when the EEPROM controller 27 detects a write operation to the EEPROM memory 26, the EEPROM_MODE register is consulted to determine whether mode “2” is selected. The elements of the data processing unit 10 involved in this activity are highlighted in FIG. 15 in heavy lines.
  • [0103]
    In one implementation, a fast write operation comprises performing a write cycle on the EEPROM for a duration that is less than the specified write cycle time recommended by the manufacturer of the particular EEPROM device being used. Typically, the logic and support circuitry for programming the EEPROM can be provided in the EEPROM controller 27. However, the programming components can be provided outside of the data processing unit 10 and controlled by the EEPROM controller via suitable control lines on the bus 14 (FIG. 1).
  • [0104]
    It is understood of course, that a fast write operation will result in data that will have a shorter retention time than data retention times from performing a full write cycle. However, in a particular application, a decreased retention time may be acceptable. In the case of a virtual machine such as the Java VM in a Javacard environment, a fast write operation of EEPROM in accordance with the present invention may provide a practical solution to the limited RAM capacity that is typical in such environments. EEPROM capacity is typically greater than RAM availability in a Javacard. Long term EEPROM retention is not needed in a situation where a VM applet is going to run once and thus only needs access to memory long enough to do its thing. The fast write operation of the present invention can therefore provide a VM applet with the memory space it needs to execute.
  • [0105]
    Another EEPROM memory access mode, mode “1” say, is provided. In this mode, all EEPROM writes are performed using standard write cycle times recommended by the manufacturer of the EEPROM device (“slow” writes, deep writes, persistent writes). In accordance with the invention, it is possible to change the EEPROM access mode dynamically to permit fast writes or slow writes as needed for any given instance. Mode “1” access comprises performing all write operations using a slow write. See FIG. 16.
  • [0106]
    In accordance with an aspect of the invention, fast and slow write operations can be performed depending on the address location. Thus, still another EEPROM memory access mode, mode “0” say, is provided. FIG. 17 illustrates a particular embodiment of this aspect of the invention. An address range EEPROM_FAST_END to EEPROMM_FAST_START is defined. In this particular embodiment, the EEPROM controller 27 is configured to recognize this range of addresses and to perform fast write operations when the addressed memory location falls within the range. Write operations for EEPROM addresses outside of this range are performed with slow write operations. In a particular implementation, the address range can be stored in registers (or some such data storage component) provided in the data processing unit 10; e.g., in the EEPROM controller 27. The address range can be hardwired values, or can be modifiable.
  • [0107]
    In another embodiment of this aspect of the invention, the EEPROM address space can be partitioned into a plurality of segment of equal or unequal sizes. Each segment can be identified as being a fast-write segment or a slow-write segment. Register or similar data storage components can be provided to identify the address ranges of each segment. Another register can be used to indicate the write-mode (fast, slow) for each segment. The EEPROM controller 27 can access the information in these registers to determine which write mode to apply for a given write operation.
  • [0108]
    Further according to the invention, the foregoing EEPROM memory access modes can be selectable at run time. Thus, a particular embodiment of a microprocessor according to the invention can provide one, two, or all three modes of operation. If two or more memory access modes are provided, each access mode can be selectable at run time. Suitable control (e.g., software) can be provided which writes an appropriate EEPROM_MODE value into the register 21 h (FIG. 2). Each application can thereby determine which EEPROM memory access policy it desires to use. In addition, the access mode can even be changed within an application whenever the application deems it is appropriate to do so.
  • [0109]
    In another embodiment of the invention mode “3” operation (caching) can be combined with operation modes 2-0. Thus, a microprocessor according to this aspect of the invention can provide multiple runtime modes of operation where an application executing on the microprocessor can dynamically switch among all the EEPROM modes of operation.
Patent Citations
Cited PatentFiling datePublication dateApplicantTitle
US4969121 *Mar 2, 1987Nov 6, 1990Altera CorporationProgrammable integrated circuit logic array device having improved microprocessor connectability
US5201044 *Apr 16, 1990Apr 6, 1993International Business Machines CorporationData processing method for file status recovery includes providing a log file of atomic transactions that may span both volatile and non volatile memory
US5325375 *Jun 28, 1991Jun 28, 1994Sun Microsystems, Inc.Method and apparatus for non-atomic level parity protection for storing data in a random access memory
US5369757 *Jun 18, 1991Nov 29, 1994Digital Equipment CorporationRecovery logging in the presence of snapshot files by ordering of buffer pool flushing
US5446862 *Nov 29, 1993Aug 29, 1995Mitsubishi Denki Kabushiki KaishaSystem and method for granting or preventing atomic or nonatomic memory access requests to shared memory regions
US5488711 *Apr 1, 1993Jan 30, 1996Microchip Technology IncorporatedSerial EEPROM device and associated method for reducing data load time using a page mode write cache
US5530958 *Dec 23, 1994Jun 25, 1996Massachusetts Institute Of TechnologyCache memory system and method with multiple hashing functions and hash control storage
US5581726 *Oct 24, 1994Dec 3, 1996Fujitsu LimitedControl system for controlling cache storage unit by using a non-volatile memory
US5696929 *Oct 3, 1995Dec 9, 1997Intel CorporationFlash EEPROM main memory in a computer system
US6075723 *Dec 14, 1998Jun 13, 2000Sony CorporationNonvolatile semiconductor memory device and IC memory card using same
US6128627 *Apr 15, 1998Oct 3, 2000Inktomi CorporationConsistent data storage in an object cache
US6272587 *May 12, 1999Aug 7, 2001Cummins Engine Company, Inc.Method and apparatus for transfer of data between cache and flash memory in an internal combustion engine control system
US6272607 *Aug 24, 1999Aug 7, 2001International Business Machines CorporationMethod and apparatus for transactional writing of data into a persistent memory
US6332185 *Jun 5, 1995Dec 18, 2001Sun Microsystems, Inc.Method and apparatus for paging data and attributes including an atomic attribute for digital data processor
US6757201 *Jan 22, 2003Jun 29, 2004Renesas Technology CorporationNonvolatile memory, IC card and data processing system
US20010023472 *Oct 16, 1998Sep 20, 2001Noriko KubushiroData storage control method and apparatus for external storage device using a plurality of flash memories
US20010047480 *Jan 5, 2001Nov 29, 2001Chiaki TanimotoIC card and microprocessor
US20020026566 *Feb 5, 1999Feb 28, 2002Kosaku AwadaData backup in non-volatile memory
US20020051394 *Aug 13, 2001May 2, 2002Tsunehiro TobitaFlash memory control method and apparatus processing system therewith
Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US7562094 *Jul 14, 2009Precise Software Solutions, Inc.Object-level database performance management
US7928763 *Jul 14, 2010Apr 19, 2011Martin VorbachMulti-core processing system
US7996827Aug 16, 2002Aug 9, 2011Martin VorbachMethod for the translation of programs for reconfigurable architectures
US8058899Feb 13, 2009Nov 15, 2011Martin VorbachLogic cell array and bus system
US8069373Jan 15, 2009Nov 29, 2011Martin VorbachMethod for debugging reconfigurable architectures
US8099618Jan 17, 2012Martin VorbachMethods and devices for treating and processing data
US8127061Feb 18, 2003Feb 28, 2012Martin VorbachBus systems and reconfiguration methods
US8145881Mar 27, 2012Martin VorbachData processing device and method
US8156284Jul 24, 2003Apr 10, 2012Martin VorbachData processing method and device
US8195856Jun 5, 2012Martin VorbachI/O and memory bus system for DFPS and units with two- or multi-dimensional programmable cell architectures
US8209653Jun 26, 2012Martin VorbachRouter
US8230411Jun 13, 2000Jul 24, 2012Martin VorbachMethod for interleaving a program over a plurality of cells
US8250503Jan 17, 2007Aug 21, 2012Martin VorbachHardware definition method including determining whether to implement a function as hardware or software
US8281108Jan 20, 2003Oct 2, 2012Martin VorbachReconfigurable general purpose processor having time restricted configurations
US8281265Nov 19, 2009Oct 2, 2012Martin VorbachMethod and device for processing data
US8301872May 4, 2005Oct 30, 2012Martin VorbachPipeline configuration protocol and configuration unit communication
US8310274Mar 4, 2011Nov 13, 2012Martin VorbachReconfigurable sequencer structure
US8312200Jul 21, 2010Nov 13, 2012Martin VorbachProcessor chip including a plurality of cache elements connected to a plurality of processor cores
US8312301Sep 30, 2009Nov 13, 2012Martin VorbachMethods and devices for treating and processing data
US8407525Mar 26, 2013Pact Xpp Technologies AgMethod for debugging reconfigurable architectures
US8429385Sep 19, 2002Apr 23, 2013Martin VorbachDevice including a field having function cells and information providing cells controlled by the function cells
US8438276May 7, 2013Precise Software Solutions, Inc.Method of monitoring network and application performance by analyzing web clients and web servers
US8468329Jun 8, 2012Jun 18, 2013Martin VorbachPipeline configuration protocol and configuration unit communication
US8471593Nov 4, 2011Jun 25, 2013Martin VorbachLogic cell array and bus system
US8686475Feb 9, 2011Apr 1, 2014Pact Xpp Technologies AgReconfigurable elements
US8686549Sep 30, 2009Apr 1, 2014Martin VorbachReconfigurable elements
US8726250Mar 10, 2010May 13, 2014Pact Xpp Technologies AgConfigurable logic integrated circuit having a multidimensional structure of configurable elements
US8788527Aug 15, 2011Jul 22, 2014Precise Software Solutions, Inc.Object-level database performance management
US8803552Sep 25, 2012Aug 12, 2014Pact Xpp Technologies AgReconfigurable sequencer structure
US8812820Feb 19, 2009Aug 19, 2014Pact Xpp Technologies AgData processing device and method
US8819505Jun 30, 2009Aug 26, 2014Pact Xpp Technologies AgData processor having disabled cores
US8869121Jul 7, 2011Oct 21, 2014Pact Xpp Technologies AgMethod for the translation of programs for reconfigurable architectures
US8914590Sep 30, 2009Dec 16, 2014Pact Xpp Technologies AgData processing method and device
US9037807Nov 11, 2010May 19, 2015Pact Xpp Technologies AgProcessor arrangement on a chip including data processing, memory, and interface elements
US9047440May 28, 2013Jun 2, 2015Pact Xpp Technologies AgLogical cell array and bus system
US9075605Oct 17, 2012Jul 7, 2015Pact Xpp Technologies AgMethods and devices for treating and processing data
US20070150637 *Feb 18, 2003Jun 28, 2007Martin VorbachBus systems and method for reconfiguration
US20090144485 *Feb 6, 2009Jun 4, 2009Martin VorbachProcess for automatic dynamic reloading of data flow processors (dfps) and units with two- or three-dimensional programmable cell architectures (fpgas, dpgas, and the like)
US20090199167 *Jan 17, 2007Aug 6, 2009Martin VorbachHardware Definition Method
US20100095088 *Sep 30, 2009Apr 15, 2010Martin VorbachReconfigurable elements
US20100122064 *Sep 30, 2009May 13, 2010Martin VorbachMethod for increasing configuration runtime of time-sliced configurations
US20100281235 *Nov 17, 2008Nov 4, 2010Martin VorbachReconfigurable floating-point and bit-level data processing unit
US20100287324 *Jul 21, 2010Nov 11, 2010Martin VorbachConfigurable logic integrated circuit having a multidimensional structure of configurable elements
US20110137717 *Dec 4, 2009Jun 9, 2011Reuthe EricSystem for Providing Digital Incentives Including a Digital Incentives Switch for Matching Transactions and Incentives
US20110161977 *Mar 23, 2010Jun 30, 2011Martin VorbachMethod and device for data processing
US20110173596 *Nov 28, 2008Jul 14, 2011Martin VorbachMethod for facilitating compilation of high-level code for varying architectures
USRE44365Oct 21, 2010Jul 9, 2013Martin VorbachMethod of self-synchronization of configurable elements of a programmable module
USRE44383Apr 24, 2008Jul 16, 2013Martin VorbachMethod of self-synchronization of configurable elements of a programmable module
USRE45109Oct 21, 2010Sep 2, 2014Pact Xpp Technologies AgMethod of self-synchronization of configurable elements of a programmable module
USRE45223Oct 21, 2010Oct 28, 2014Pact Xpp Technologies AgMethod of self-synchronization of configurable elements of a programmable module
Classifications
U.S. Classification711/202, 712/E09.033, 711/E12.014, 711/E12.102, 711/E12.017
International ClassificationG06F12/06, G06F12/10, G06F12/02, G06F9/44, G06F9/312, G06F12/14, G06F12/08
Cooperative ClassificationG06F12/0802, G06F12/0292, G06F12/145, G06F9/30043
European ClassificationG06F9/30A2L, G06F12/14C2, G06F12/08B, G06F12/02D6
Legal Events
DateCodeEventDescription
Oct 28, 2003ASAssignment
Owner name: RENESAS TECHNOLOGY AMERICA, INC., CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:MORITA, TOSHIYASU;KAWASAKI, SHUMPEI;REEL/FRAME:014670/0799
Effective date: 20031017
Jul 28, 2011ASAssignment
Owner name: RENESAS ELECTRONICS AMERICA INC., CALIFORNIA
Free format text: MERGER;ASSIGNOR:RENESAS TECHNOLOGY AMERICA, INC.;REEL/FRAME:026669/0788
Effective date: 20100401
Sep 21, 2011FPAYFee payment
Year of fee payment: 4
Jan 27, 2016FPAYFee payment
Year of fee payment: 8